Issue with mouse picking on PolyData line structure

[tamaynard]

I'm visualizing a line-based dataset consisting of a sequence of point data. The lines are rendering correctly, however it seems the world space coordinates (worldPosition) returned by the mouse picking feature are incorrect. The value returned appears dependent on the camera position, whereas I would expect the result to be consistent and in line with the input data no matter where the camera is located. I've attached two example files, one an overwritten OBJViewer.js example modified to display lines based on point data contained within example_data.js. To reproduce simply paste these in the usage/Geometry folder and run the build/serve scripts.
Geometry.zip

[tamaynard]

It appears the Z value is relative to the camera. Is there a function already implemented to convert this to world coordinates, or should I just write one myself?

[tamaynard]

Probably related to the HardwareSelector's captureZValues parameter.

[jourdain]

1.2.1 should have tacking care of it. Which version are you using?

[tamaynard]

I'm using the current release 1.8.2, just double checked to be certain.

[jourdain]

Could this be a HiDPI issue with invalid xy input?

[tamaynard]

I don't think so, I'm using standard 1080p monitors on both my desktop and laptop, and I can reproduce it on both systems.

[jourdain]

Thanks for cross checking those ahead of a deeper investigation. ;-)

[tamaynard]

No problem, appreciate any help you can provide as this is the last remaining (known) issue holding up deployment of a prototype of the app I'm developing for a client.

[floryst]

Hi @tamaynard, could you provide a minimal repro of the issue using this codesandbox template? That will help us determine the root cause of your issue.

[tamaynard]

Here you go @floryst. To see the issue, simply click around the vertical line and notice the incorrect Z-value in the debug output on the left-hand side of the screen. Zooming closer to the line using the mousewheel will result in smaller Z-values (which is why I'm thinking it's using the camera's distance from the object as the Z-value).

[jourdain]

Just edited the code to have a cube as it is easier to validate since each face should have a fix value along 1 axis.

https://codesandbox.io/s/react-vtk-js-base-forked-0hb5y?file=/src/App.js

[tamaynard]

Not quite as prominent, but yes the values don't appear to be correct with the cube either @jourdain. I never noticed until attempting to pick the rendered line as in my example; there the values were completely out of left field so it stuck out like a sore thumb.

[tamaynard]

Has there been any movement on this?

[jourdain]

I don't think so. @martinken do you see anything strange in that method?

You can use that example to see that the returned world coordinates are off as we can expect the faces of a cube to have a fix component value.

[martinken]

I'll take a look at this right now and see what's going on.